Latest / The Diary of a CFO / Building an AI-Ready Finance Team Without Losing the Human Side, with Tariq Munir

Transcript

Building an AI-Ready Finance Team Without Losing the Human Side, with Tariq Munir · The Diary of a CFO

0:00
Hello and welcome to the Diary of a CFO podcast.
0:02
I'm your host with Wassia Kamon. I'm a CFO with
0:05
a background in accounting, FBNA, and I started
0:07
this show to talk about what leading in finance
0:10
really looks like and what it takes to become
0:12
a CFO. Each week, we explore how today's top
0:14
finance leaders build high -performing teams,
0:16
partner with CEOs and boards, and lead through
0:18
growth and transformation without burning out
0:21
in the process. Today, I'm super delighted to
0:23
have with me Tariq Munir. He is a keynote speaker
0:26
and digital transformation and AI advisor. with
0:29
over 20 years of experience guiding Fortune 500
0:32
leaders through organizational change. He's also
0:35
a LinkedIn top voice and the author of Reimagine
0:37
Finance. It's a book that explores how finance
0:40
leaders can navigate disruption while amplifying
0:43
uniquely human capabilities. Welcome to the show,
0:46
Derek. Thank you, Vasya, for having me. Super
0:48
excited to be here. Oh, same here, because I
0:51
really want us to be able to talk about how to
0:53
build an AI finance, AI ready finance team in
0:57
12 months without losing that human side, which
0:59
is really what I really liked about the angle
1:01
you took in your book. So I'm curious to understand
1:03
from you when you say AI ready finance team,
1:07
what does that actually look like in reality
1:10
day to day? So a great question was, yeah. And
1:13
well, you know, theoretically, We can do a whole
1:16
lot of readiness assessments. We can do all those
1:19
pre -assignment or pre -fillout checklists, do
1:24
all those sort of questionnaires. However, I
1:28
always say that look at just one day in practice
1:31
that can actually give you a good litmus test
1:35
if you are ready to have AI in your business
1:40
or not. And that one litmus test is understanding
1:45
or evaluating how are you making your decisions
1:47
today. If data is just another chart on a PowerPoint
1:53
deck, and that is it, and does not fundamentally
1:56
drive how you are making decisions, I'm sorry,
1:59
you are not ready. Most of us think that once
2:03
we get AI, we will become data driven because
2:06
then we'll have more insights, faster insights,
2:09
and so on and so forth. It is in fact the other
2:11
way around. To really leverage AI, we need to
2:16
have a data -driven mindset. Because at the end
2:18
of the day, data -driven is not a technology
2:20
issue. It's a mindset thing. We need to be able
2:24
to use the output of machines, rely on the output
2:27
of the machine. And to do that, we need to be
2:30
able to have that courage to take a decision
2:33
based on data. Okay, while our gut or instinct
2:37
tells us otherwise because a lot of times we
2:40
go with our gut We go with our instinct and I
2:42
don't say that, you know, don't go with your
2:43
gut or instinct. That's what make us you makes
2:45
us uniquely human though but Relying on data
2:49
when data is telling us an entirely different
2:51
story is what makes us ready to go for any technology
2:55
for that matter Oh, I would like to see how it
2:58
works in real life, right? So let's say you walk
3:01
into a finance department today. What are the
3:03
signal that will tell you they're not ready for
3:05
AI, right? And that's even if, like you said,
3:08
the technology out there. So I walk into a finance
3:10
team and what tells you that they're not data
3:13
driven enough already to be ready for AI? So
3:17
as I mentioned, Vasya, if most of the decisions
3:20
are still based on gut feel or the person, instead
3:24
of the merit or what the data is telling us,
3:28
they are based on the hierarchy of decision -making
3:32
or hierarchy of people in the organization, we're
3:36
not data -driven then. We are more people -driven.
3:40
Not sorry, people -driven, I would say we are
3:42
more... guided by the instinct and gut and our
3:48
experience as opposed to what data is telling
3:52
us. Apart from that, I would like to point out
3:55
two more things. One is, of course, being data
3:58
-driven. Secondly, it's about workflows being
4:01
too complex. If I walk into a finance function
4:05
and their processes are broken, their processes
4:07
are not supporting automation, there are too
4:10
many manual workarounds, Manual workarounds is
4:15
not a bad thing per se, but doing it in the form
4:19
of something like creating a manual workaround
4:23
to overcome some system issue is what I'm talking
4:28
about. So if those workflows are too complex,
4:31
all we are going to do using AI is amplify that
4:34
complexity. So I always start off with streamlining
4:40
those workflows. And lastly, a very important
4:44
element I always recommend and advise my clients
4:47
around is that if people are hesitant to experiment,
4:52
everything needs to be safe. Everything needs
4:53
to be risk -averse. Being finance, we are risk
4:56
-averse. That's how we are trained, right? But
5:00
AI on the other hand requires some experimentation.
5:03
It's not a kind of an implementation which will
5:05
happen like an ERP or any other cloud or SAS
5:07
solution where there'll be clear go, no go, no
5:10
go decisions. And you will always have an output
5:13
which is not very accurate. Then you will retrain
5:16
your models to get more accurate output and so
5:18
on and so forth. So you need to have that experimentation
5:22
mindset. Now these are the things which I'm talking
5:24
about right now does not really need technology
5:27
at all. as you would have noticed, right? These
5:29
are mostly mindset -driven. These are mostly
5:32
behavior -driven things, which would actually
5:34
drive and get you ready to leverage AI. And I
5:38
also don't say that you do all of these things
5:40
and then go for AI. Of course, that's also not
5:43
practically possible. But starting to realize
5:46
and starting to acknowledge that we need to change
5:49
is the first step. And then working on this while
5:52
you are, of course, working on your AI and digital
5:54
roadmaps as well is what sets us up for success.
5:58
Okay. So if I understand correctly, we start
6:00
with the idea of we need to be more data -driven.
6:03
So let's say I'm running a forecast and for the
6:06
past three years, I see that sales have been
6:09
increasing at 5%. If it's data -driven, when
6:11
I'm looking at my next year, I should, you know,
6:14
also feel like it's going up. for example, but
6:18
let's say the marketing director comes in and
6:20
say, oh no, let's lower it, right? So now I'm
6:23
almost not as data driven, but now it gets into
6:26
the politics. Is that correct? Yeah, yeah, yeah.
6:29
You are absolutely, you are absolutely nailing
6:31
it. And that's where the biggest issue comes
6:34
in as well, right? I mean, if market is sending
6:37
us different signals, But because of our sales,
6:41
our marketing directors based on their, or even
6:43
CFOs for that matter, based on their experience,
6:45
they believe that this is something which is
6:48
temporary or, you know, it's fine. We'll be able
6:50
to get over with that. Or we have done that in
6:52
the past and we know how it's going to work.
6:55
So if we ignore what data is telling us, then
6:58
of course we are not data driven. Having said
7:00
that, there is always that balance between the
7:02
two. Yes, as a human, we must use our experience.
7:05
This is what makes us at the end of the uniquely
7:07
human. What I'm saying is that if data is telling
7:09
us an entirely different story and our gut instinct
7:12
is telling us an entirely different story, we
7:14
need to really critically evaluate that situation
7:17
as opposed to just, you know, saying data is
7:19
fine, data is not right. Or, you know, most of
7:21
the time you might have heard in a boardroom
7:23
or in those those planning meetings, oh, I don't
7:26
trust the number, I don't know if this number
7:28
is correct or not, and so on and so forth. So
7:31
again, there are a lot of different elements
7:32
associated with building that trust as well.
7:36
But inherently not trusting data or inherently
7:38
having that outset attitude towards not relying
7:43
on data for decision making is how we drop the
7:48
ball on being data driven. Okay, and then I understand
7:51
the part about the processes where you have to
7:54
do patchwork So if you used to do it that I can
7:57
imagine how it will be hard to automate those
8:00
kind of things with technology, right? You have
8:03
to fix how you're doing things first. But then
8:05
when you talk about that mindset of experimenting,
8:08
like you said, in finance, it's hard to just
8:11
experiment because it's like compliance is reporting
8:14
to the board. If there is an hallucination or
8:17
something like that, like, how do you overcome
8:20
that? How do you prepare that mindset to be willing
8:24
to experiment and start moving toward AI readiness
8:28
in finance? Great question, Vasia. And whenever
8:31
I say experimentation, we need to step back a
8:35
little bit. OK. Experimentation is one thing
8:39
and making mistakes is another thing. Right.
8:43
If we are. We are trying to do something say
8:47
for example I get a chat dpt or some enterprise
8:50
version and I'm just using it to it to as an
8:54
experiment in my mind I am just using it to create
8:57
a board report and then I just send it through
8:59
to To board of directors. That's not an experiment.
9:03
That's a mistake Right. So so so we need to very
9:08
clearly understand that difference between experimentation
9:11
and mistakes number one number two Experimentation
9:15
is not something we just do like, I mean, we're
9:18
not sitting in a lab and we are just trying to,
9:20
even scientists don't do experiments by that,
9:23
right? By that, that they just start combining
9:25
two things and think about, okay, something comes
9:27
up. Combine tomatoes and oranges. Yeah, right,
9:29
something comes up. Experimentation, what I always
9:34
say that you need to earn the right to experiment.
9:38
And how you do that is by clearly understanding
9:41
first of all, what are your clear business needs?
9:44
Number one, what are the problems that you are
9:46
trying to solve and then finding a different
9:49
solutions against those problems and then based
9:52
on the impact value complexity, identifying those
9:56
top two, three candidates or one candidate where
9:59
you can actually go. in and start to build those
10:03
experimentation, build those experiments and
10:05
do experiments around that. It does not mean
10:08
that you just go in blindly and start using AI
10:13
to create output and then in the name of experimentation,
10:16
you are submitting a wrong reports for the compliance
10:19
or doing doing other or exposing organization
10:23
to the risks. It should not be exposed. So that's
10:26
that's not experimentation again, you know, that's
10:28
that's not not not how we do it. We need to be
10:31
thoughtful about our experiments. And then there
10:34
are certain core elements. So you need to do
10:37
your text submissions, for example. You need
10:39
to do your SEC submissions. There is no room
10:43
for us to experiment on the output. So output
10:47
needs to be clear. But how we generate the input,
10:50
we can always experiment. We can always try to
10:53
find different ways or better ways of doing things.
10:55
But that does not give me any right. to not validate
11:00
or verify the output and just send it through
11:02
to the board or to the regulator. So experimentation
11:11
needs to be thoughtful, needs to be very clearly
11:14
tied into your specific business need. It's not
11:18
about that just because we want to use a native
11:21
AI somehow. So let's do an experiment around
11:25
that. That's not how we do experimentation. So
11:27
that needs to be that distinction needs to be
11:30
very, very clear. And I agree. I mean, sometimes
11:33
we take that as our experimentation as a wrong
11:36
in a wrong connotation and think about it more
11:39
like, in fact, what we are doing is not experiments.
11:43
What we are doing are making mistakes. So we
11:45
need to make sure that we bifurcate and we separate
11:50
the two. I hope that explains a bit of that question.
11:53
Oh yeah, absolutely. And I'm curious to hear
11:55
from your work, what have been some of those
11:58
mistakes that you have seen? Because clearly
12:00
one of them is just not double checking the work.
12:02
Like we do it in accounting, internal controls,
12:05
right? View the work or you sign a journal entry.
12:07
But what are other mistakes that you've seen
12:09
people make as they're trying to be ready for
12:12
AI or actually implementing AI within their organizations?
12:18
I think the biggest mistake and sometimes I call
12:22
it bit of a pandemic kind of a situation is becoming
12:26
too much tool or AI focused, to be honest, and
12:31
not actually thinking about the business problem.
12:34
The very essence of experimentation, what I was
12:36
talking about, if we are ignoring that, that's
12:38
the biggest mistake in itself. So what happens
12:41
in various scenarios? I have seen that. We start
12:46
the conversation, so I approach or someone approaches
12:49
me around AI transformation or this kind of work
12:53
that I do. Many times the conversation unfortunately
12:57
starts with, how can I automate this task? How
12:59
can I use AI to do this thing better? That's
13:02
a wrong question to ask. that I would sometimes
13:05
say as close as doing a mistake because what
13:07
we do is then when we start finding point solutions
13:11
which might give us some productivity benefits
13:14
or might you know automate some of the task but
13:17
at that very specific boundary within which that
13:20
AI would be operating. AI's real value comes
13:22
in when it integrates across the organization,
13:25
when it fits into your broader vision, your broader
13:28
organizational operating model for that matter.
13:32
That is, I believe, the biggest mistake that
13:34
we make when we are. I have seen AI projects
13:41
ranging from cash flow forecasting to revenue
13:45
analytics to process optimizations not performing
13:49
well or not doing as planned because there was
13:53
nothing wrong with the approach. There was nothing
13:56
wrong with the tool or the technology itself
13:59
or its its appropriateness for that problem,
14:06
but it didn't fit into the broader organizational
14:09
operating model or the broader organizational
14:10
structure. People did not rely on the output.
14:14
At the end of the day, people did not adopt the
14:16
technology. I mean, having a tool and then using
14:18
that tool are two different things, right? I
14:21
mean, adopting and deployment. So a lot of times,
14:25
these are the kind of, I would say, mistakes
14:28
or or not the right way of doing experimentation
14:32
that I have witnessed quite a few times. Quite
14:34
a few times. Sometimes I sound more like a broken
14:37
record as well when I keep telling people that,
14:39
you know, please let's be a little bit more thoughtful
14:43
around our experimentation. Let's not talk about
14:45
tool first. Let's not talk about the task even.
14:48
Let's look at the holistic system. What are we
14:51
doing, right? I mean, how our work is impacting.
14:54
At the end of the day, remember, finance is up.
14:58
is a function of every other function in the
15:01
organization. We cannot say that even the basic
15:06
accounting work that we are doing can happen
15:08
in isolation without the input from commercials,
15:12
from operations, from cost of goods sold, for
15:15
sales from the sales team. It cannot happen.
15:18
How that linkage, how that end -to -end workflow,
15:22
the end -to -end processes work in order for
15:25
us to get the maximum utilization of AI or the
15:29
maximum leverage out of AI. So that is what I
15:33
see a lot in the work that I do. OK. And so I'm
15:37
curious too, in your work and your book, you
15:39
talk a lot about human centric enabled finance.
15:43
So what needs to change first in how finance
15:47
people think or behave for AI to work in our
15:51
favor? Like, what do we need to change? Because
15:53
I know you said the first question sometimes
15:55
is, how can I use it to automate this? Well,
15:59
I'm trying to make my life easy, so I'll likely
16:01
ask the same question. So what is a better question
16:04
to ask? What actually needs to change for finance
16:07
to be more ready for AI? I believe the fundamental
16:10
change which is needed in any finance function
16:14
or for that matter, I mean broader organization
16:17
because I do work with broader organization as
16:19
well. So is how do we build a digital mindset
16:23
within the team? Now, again, as I said, I sometimes
16:26
sound like a broken record when I say that digital
16:29
mindset is a very simple thing. It is how well
16:33
the finance team or business team is able to
16:36
utilize the output of machine to solve a real
16:40
world business problem. That is it. As long as
16:42
you are able to connect the dots between the
16:44
two, you have a digital mindset and what that
16:47
enables us to become what we are supposed to
16:51
do. We become better business partner and we
16:53
actually navigate the ship as opposed to just
16:56
being those stewards of the data. When we build
16:59
that digital mindset, two things happen. One,
17:04
by default, we get process efficiencies, faster
17:07
insights, of course, because now you are able
17:09
to connect that what are the different business
17:12
problems I'm facing. And how do I use technology
17:15
holistically to solve those business problems?
17:19
So as an output, you get process efficiencies,
17:24
you get productivity, you get faster insight.
17:26
And what that does, secondly, and most important,
17:30
I would say, this enables us to do what we are
17:36
designed to do as humans, what we are meant to
17:39
do. critical thinking business partnering and
17:42
not that manual tedious manual work that we are
17:45
we are we we spend so much time on so as as humans
17:48
I think we can do much better than what we are
17:51
doing today as finance leaders we can do much
17:53
better than how we are we are operate making
17:56
our business plans or how we are doing our back
17:59
office accounting for that matter or how we are
18:02
paying invoices to our suppliers and so on and
18:05
so forth. We can do a much, much better job.
18:07
I'm not saying that we're not doing a great job.
18:09
I mean, finance is one of the most intelligent
18:12
teams in the organization, and they are one of
18:14
the most hardworking as well. But how do we make
18:17
them a little bit more smart workers using technology?
18:21
And this is what helps us then build a human
18:25
-centric tech -enabled finance function. And
18:29
so building that digital mindset, when you think
18:32
about building that digital mindset, what a practical
18:35
step to get there, right? Like, because I know
18:38
it's a bit out there. I see the bigger picture.
18:42
But for somebody listening, I want to build a
18:44
digital mindset within my team. As a leader,
18:47
what do I need to do? And what does my team need
18:50
to do? What do I need to expose them to in order
18:52
to really elevate that mindset? Great question.
18:56
And I actually talk in quite a detail in my book
18:59
as well, specifically on this building a digital
19:02
mindset. So my argument is that we need to build
19:04
two things. First of all, is your digital literacy,
19:08
of course, right? Now, that does not mean that
19:10
you need to become a data scientist or you need
19:12
to become an engineer unless you want to, of
19:15
course. But you need to understand that how different
19:19
technologies that exist today, what can we Use
19:23
to solve some of the business problems today.
19:25
What are the trends? Read a lot of books. I always
19:28
recommend that there is no substitute to reading
19:30
good books. Understand there are a lot of free
19:34
courses available out there where you can just
19:36
get an understanding of what AI is all about,
19:39
how it can help us solve some of those pressing
19:41
issues that we face today. What are the different?
19:45
different methods within data analytics that
19:48
we can apply and use them, and depending upon,
19:51
of course, where you are in your hierarchy in
19:53
the organization, number one. And number two,
19:57
by building a digital mindset requires us to
19:59
have a growth mindset. So it is almost like a
20:03
combination of the two, digital literacy and
20:05
growth mindset gives you a digital mindset. Growth
20:07
mindset means that as a human, and that applies
20:12
of course to both divided teams, but sometimes
20:16
I say that it applies more to the leaders in
20:19
the organization because sometimes that's where
20:22
people look up to and follow the lead and follow
20:25
their footsteps. If as leaders we have a growth
20:29
mindset, we believe that as humans we can learn
20:32
any skill, there is no limit. to human mind on
20:40
what can be taught or what can be learned, we
20:43
build that growth mindset and that requires us
20:45
to do continuous learning, understand different
20:48
aspects. And first of all, be very open and acknowledging
20:52
of the fact that, yes, I mean, I don't have answers
20:55
to everything, but I'm willing to go out and
20:58
I'm willing to answer, to find those answers.
21:01
When you create that kind of a culture as a leader,
21:05
within your team, your team follows through your
21:08
footsteps and then they also work towards finding
21:12
those answers. And when you are trying to find
21:13
those answers, you will, again, I mean, there
21:16
is no simple one, one tool or one course available
21:21
out there or a magic training available out there
21:25
that I can just tell you to do it. And it will
21:27
get, it will, it will, you know, help you become
21:29
a digital mindset. It's a continuous learning
21:31
journey and the continuous learning as I always
21:34
say is is not about how many trainings you are
21:36
doing, how many hours you are training, hours
21:40
are you clocking or how many certificates you
21:42
are getting. It is about acknowledging that it
21:47
is about the willingness to challenge your very
21:52
own assumptions in the light of new knowledge.
21:55
That's what growth mindset is all about. That's
21:58
what continuous learning is all about. So it's
22:00
all behavioral thing. It has to be ingrained
22:03
within our DNA. We need to ingrain this within
22:06
our DNA in order for us to be able to build that
22:10
digital mindset eventually. And it won't happen
22:12
overnight, but we have to take the first step
22:15
towards it. Yeah, thank you so much for defining
22:18
and I'm curious, you know, I understand and I've
22:23
seen how inspiring it is for a team to see their
22:27
leader. learn something new and share it with
22:30
them during meetings. And they're like, OK, so
22:32
they're coming up with this. I probably need
22:34
to catch up. And I can see how it will contaminate,
22:37
quote, unquote, the rest of the team to kind
22:40
of elevate their game. So as we look further
22:43
and we look further out, if you had to list the
22:46
key skills that an AI ready finance professional
22:49
needs to develop over the next few years, what
22:52
will be on that short list? Of course, the first
22:55
one will be that digital mindset, right? As you
22:58
always say, but I think as as as future finance
23:01
leaders or as future business leader for that
23:03
matter, there is one thing which is common, which
23:06
is going to be the common through line, which
23:08
is actually the common through line at this stage
23:10
as well. And that is we are all change managers
23:13
of the future. Change is the is the is the I
23:15
mean, it's a cliche that change is the only constant.
23:18
Yeah. So we need to become excellent at managing
23:23
change. which of course requires us to systematically
23:26
understand what are the business problems we
23:28
are facing, what is in it for people, how we
23:31
keep them engaged, how we get them on board with
23:34
the change. So first of all, we need to build
23:37
our skills around becoming a better change managers.
23:41
With AI coming in, there is going to be a lot
23:47
of productivity. that comes out of course there
23:51
is a lot that happens as a as as part of the
23:54
output of technology how that technology works
23:57
and what it does is still kind of a black box
24:01
and that is where the skill set around responsible
24:04
and ethical AI becomes the one of the most important
24:08
ones in future as finance leaders we are yes
24:13
we and we are in in a bit of um advantageous
24:16
position, I would say, because we have a bit
24:18
of a head start with our training and our predisposition
24:23
to risk and governance and control. We can actually
24:27
lead the charge in this responsible and ethical
24:30
AI. As CFOs, we can ask the question that why
24:33
we should be doing something. What are the consequences
24:36
of doing that? How data is currently flowing
24:39
through within the organization? Having those
24:42
right. making those right investment calls and
24:45
right ethical decisions. Again, at the end of
24:49
the day, responsible and ethical AI is a proportionate
24:51
ownership. It's not something that only CFO can
24:54
do or CSO can do or CTO can do. But as a CFOs,
24:57
as a finance leaders, due to our position within
25:00
the organization and how people look at to us
25:03
for the ethical land and governance decisions,
25:06
we can influence that a lot. So that is one skill
25:09
set, which I always recommend a lot of finance,
25:12
which I always recommend to finance leaders to
25:14
definitely get ahead on. get understanding of
25:18
various aspects around what are the different
25:20
regulations coming in, what are the different
25:23
policies, procedures, and recommendations that
25:26
come from different organizations around global
25:28
organizations like UNESCO and OECD around responsible
25:32
and ethical AI. Number two. And then I think
25:36
another important skill that we need is When
25:42
AI is determining the most of the output and
25:45
now we are talking about agentic AI, which is
25:47
actually have that agency and doing stuff for
25:50
us. A lot of times we get there is this curtain
25:56
of technology or curtain of AI that sits between
26:01
us and the humans on the other side. So understanding
26:05
the emotions of humans within the loop, within
26:10
that orchestration of agentic AI and all the
26:13
technological transformation is the key aspect
26:17
and what we can also call emotional intelligence.
26:21
That is going to be one of the most, and in fact,
26:24
that is one of the most important skill set to
26:27
have and to build with us in order for us to
26:29
be able to, because again, when we talk about
26:31
technology, our first reaction is, oh, it is
26:35
going to take away jobs. Yes. We forget that
26:39
jobs or roles is just a description on a piece
26:45
of paper. There are real humans who are working
26:50
around those tasks and roles. What are their
26:53
emotions? Being aware of what are their training
26:57
needs? How does our decisions are impacting the
27:01
people? They are not just the numbers or they
27:05
are not just a cog in a machine. They are real
27:09
beings, conscious beings who have. emotions who
27:13
have social structures who have social needs
27:15
so how do we how do we remain aware of their
27:17
needs when the technology comes in and take away
27:20
and automate roles or automate tasks how do we
27:24
separate humans from tasks and how do we understand
27:27
their needs and how do we actually help them
27:29
grow as well while while of course making some
27:32
tough calls as well and they will be tough right
27:34
i don't hide away from the fact that roles are
27:37
not gonna get impacted they will but how we manage
27:41
those Again, goes back to change management as
27:43
well as being emotionally aware is a very critical
27:47
skill set that we need. Wow. Thank you so much.
27:51
That was great because you started with, you
27:53
know, when we talk about the skills that people
27:55
will need to develop, digital mindset, definitely
27:58
change management, having emotional intelligence
28:02
and also the idea of responsible and ethical
28:06
AI, which that's a part. I don't think we talk.
28:09
a lot about. So when we think about responsible
28:13
AI, what are some examples of cases maybe where
28:17
people used AI in no ethical way, but it made
28:21
business sense or it was a good shortcut or productivity
28:23
that happened? And what should people really
28:26
be aware when it comes to using AI from a governance
28:30
and regulation standpoint? You are absolutely
28:33
right. First of all, that we don't talk about
28:36
it a lot. right and the reason being the technology
28:40
takes precedence business use business case takes
28:43
precedence and then governance and these ethical
28:46
issues or responsibly it comes as a as a second
28:49
thought we have seen a lot of examples which
28:52
are like pretty available published as well examples
28:55
out there where it made a business sense but
28:58
it was not ethically ethically right to do so
29:01
For example, first of all, before I give an example
29:05
of or a case study around that, if we have data
29:10
from our customers and we are just using it,
29:14
we are just getting that data while we are, of
29:17
course, selling them or are transacting with
29:19
them. Just because we have that data does not
29:22
mean we own it or we can use it to train AI models.
29:27
We need explicit consent from them. That is where
29:30
these ethical things, ethical dilemmas comes
29:32
into play. We need to be very 100 % sure that,
29:35
yes, our customers, our consumers are willing
29:38
to give us that data. They have already given
29:41
us that data, but they are also willing for us
29:44
to use that data to train our algorithms. So
29:48
for example, I give this example of EverApp.
29:52
There was this this app, which was something
29:55
like Google Drive, where people would upload
29:57
their photos and save their photos. So what they
30:01
were doing, they were using those photos or people's
30:04
pictures to train some of the algorithms on facial
30:08
recognition, and then were marketing and selling
30:10
those algorithms to the other companies. So FTC,
30:14
Federal Trade Commission, they imposed on them
30:16
a penalty called algorithmic disgorgement. So
30:21
one thing is ethical, but the other thing is
30:23
the regulatory impact as well and getting in
30:25
the public eye as well. What that meant for them
30:28
was they had to destroy all their algorithms
30:32
and had to pay fines and so on and so forth.
30:36
So again, we should not be ethical because Just
30:42
we should not just be only be ethical because
30:44
if we have a if we have a If we if someone if
30:48
we have a regulator sitting on top of us, of
30:51
course, we should always be ethical But there
30:54
are of course those consequences as well secondly
30:57
one of the very important aspect that comes out
31:02
of AI and and sometimes we we see that a lot
31:05
and a output comes from AI and we say that oh
31:10
This is the output from the machine. This is
31:12
the output from the AI. So as a human, I'm not
31:16
responsible for it. However, this is the basic
31:21
premise of responsible and ethical AI that any
31:26
output by machine is the responsibility of AI
31:30
or is the responsibility of a human at the end
31:33
of the day. We are responsible for the output
31:35
of the AI machines are not so taking that ownership
31:39
is critical and that is where where that proportionate
31:43
ownership comes into play or CFOs role also can
31:46
can play up CFOs can play a bigger role in enabling
31:48
that to happen that yes we are saying so I always
31:52
say that use up what I what I advise is that
31:56
whenever you have these agentic workflows of
31:58
today or a eyes of today follow a no offense
32:01
policy every AI must have an honor and that honor
32:06
should be responsible for the output of the AI.
32:08
If something is going wrong, that owner, they
32:11
should be raising the, they should be raising
32:13
an alarm. They should be raising their hand and
32:15
saying, this is not working or this is how it
32:17
is. So, I mean, again, it's a complex topic.
32:21
It is, it is an, and I would say it's an evolving
32:23
field because we, what happened, unfortunately,
32:27
I would say, or whatever, for whatever reasons,
32:30
commercial reasons or, or, or, or corporate reasons,
32:34
we build technology and we did not think about
32:37
responsibility and ethics. till a later stage.
32:40
So even now we are building technology at a much
32:43
faster pace than responsibility or ethical practices
32:46
can actually catch up. Wow. Or sometimes I say
32:51
that we are, a lot of corporations today are
32:54
building technologies around AI and responsible,
32:58
sorry, building technologies on around AI and
33:01
are expecting someone else to do those regulations
33:05
and ethics. So they, their role becomes more
33:08
about innovation and they expect governments
33:11
or other regulators to come in and build responsible
33:15
and ethical AI frameworks. That is not how it
33:18
will work or that is not how it is sustainable.
33:21
Right from the beginning, right when we are writing
33:24
the first code for any algorithm, we need to
33:28
have responsible and ethical frameworks in place,
33:32
which unfortunately is not. that prevalent at
33:35
this stage. But we are moving in that direction.
33:38
There are people and organizations who are actually
33:41
trying to get ahead of this innovation spree,
33:47
if we call it. Wow. But there is so much pressure,
33:50
though, and also so much money to be made, right?
33:53
So that's usually what gets us in trouble. Yeah,
33:57
that's the reason, right? I mean, that's what
33:59
I was talking about, that the commercial reasons
34:00
have made it much more faster. deploy and then
34:05
just you start using AI without understanding
34:08
what are the responsible ethical social implications
34:10
of those right I mean we don't know today that
34:13
the kids that are using we didn't know about
34:16
social media how it's gonna impact us we don't
34:19
have a clue what AI is going to do for us so
34:22
going to do with our brains going to do with
34:25
our with the fit how in a society we are operating
34:27
as as a human as a human society, the way we
34:32
are operating right now, what AI is going to
34:35
do with that. We don't know because we haven't
34:37
really thought about it. It is coming as an afterthought
34:40
once everything, everyone has started to actually
34:43
use AI, which is scary as well. It is. It is.
34:47
It's very scary. And I'm curious to hear from
34:50
your standpoint as you think about the current
34:52
organization, the current structure or org chart
34:56
of a finance department, what do you see as roles
34:59
that will disappear, how some of the current
35:02
roles will evolve, like who will be part, besides
35:06
AI agent, because now I'm getting used to the
35:09
idea that it will be AI agent on, you know, on
35:11
finance teams, like, beyond that, like, how are
35:15
those roles that we're used to going to evolve?
35:18
It's, it's the, again, first of all, the, the,
35:22
the, the org chart is, is one thing is for sure
35:27
that org chart is going to change. Right? It
35:29
is and it is already already we see that happening
35:33
in many organizations today. Now, many of the
35:39
roles might not even exist today. So, for example,
35:44
when AI agents as we talked about when they are
35:46
carrying out the actual execution solving problems,
35:50
we will need humans in the loop. Right? And these
35:54
could be your FPNA AI controllers, your AI agent
35:58
coordinators, your AI controllers. Again, it's
36:02
very hard to say what they will look like, what
36:05
those role will exactly look like. And I acknowledge
36:07
that we don't know 100%. We can think of having
36:13
some of those humans in the loop within that
36:15
structure, of course, definitely. But again,
36:18
we don't know clearly at this stage that how
36:20
that exact organization are going to look like.
36:24
Having said that, of course, there will be roles
36:25
that will disappear. those roles that are that
36:30
have some problems even even the rules which
36:32
require judgment and some problem solving agents
36:35
can actually do that quite quite diligently and
36:39
in a quite efficient way, but we will have a
36:42
lot of new roles as I mentioned coming up which
36:45
will help us to ensure that a agents are performing
36:48
their tasks correctly or Identifying their objectives
36:52
or defining their objectives or what problems
36:55
they are going to solve Wow Now irrespective
36:59
of whatever happens in that scenario how that
37:02
future org structure looks like it's evolving
37:05
at this stage So it's hard to say that whatever
37:07
it will exactly look like one role or one skill
37:11
is going to Stay and that is I think is is going
37:16
to become the most relevant for anyone in the
37:20
org structure And that is the critical thinking
37:22
piece With AI, we have too many answers. True.
37:28
You ask something, you just ask a simple anything,
37:31
you know, do this or do that. It'll give you
37:33
a page full of recommendations and stuff. The
37:36
most important skill is the ability to ask the
37:40
right questions. As humans or as finance leaders,
37:45
we will need to be able to identify or define
37:51
those objectives. or the right, we cannot just
37:54
say to AI agents or just do this, process this
37:57
invoice. We need to be able to ask the right
38:00
questions. We need to be able to define the right
38:02
objectives for us to be able to leverage the
38:06
maximum out of these machines. Why an output
38:08
of machine is what it is. We should be able to
38:10
ask those questions. We cannot just rely on the
38:12
output of machine, become victim of automation
38:15
bias, because if it's created by machine, it
38:18
must be correct. No, we cannot do that. We have
38:21
to have that critical mindset, that critical
38:23
thinking, that why should we even rely on the
38:25
output of a machine? This ability to challenge
38:28
the status quo, I think, is the single most important
38:32
skill or the role that humans are going to play
38:35
in that new org chart. So everyone is going to
38:39
become that change manager, that critical thinker,
38:43
for them to be able to orchestrate orchestrate
38:47
these different machines or these different agents
38:49
together to solve the business problems that
38:51
we are trying to solve. Yeah, and I can see how
38:54
you have to have that critical thinking to ask
38:56
the right question, but also validate, like you
38:59
said, whatever the output that you needed on
39:02
both sides from end to end, which is quite. Quite
39:06
fascinating, but I'm curious to also hear, what
39:09
are some AI cases that you think CFOs should
39:13
try? Or maybe a use case where it sounded great
39:17
and it didn't go so well. I'm curious if you
39:19
have any success stories or any oops, it didn't
39:23
work out as we planned stories. Yeah, so again,
39:26
I mean, to answer your first part of the question,
39:29
the art. What are some of the use cases that
39:32
a let's see if we should try now I always say
39:35
that it all depends on first of all the business
39:38
needs Right. What is our strategy? How we are
39:41
what is our broader vision that looks like? So
39:44
so I mean rather than going a use case approach
39:47
or point solutions look at the bigger picture
39:49
But as a as a principle as a principle I would
39:53
recommend go for the well -established and mature
39:56
solutions in the beginning You don't need to,
39:58
yes, AI agents or generative AI looks very fancy
40:02
on the paper, but technology is still evolving.
40:05
It is untested in many business scenarios. We
40:07
have seen it not scaling in MIT's report, that
40:11
famous report that came up a couple of months
40:13
ago that 95 % of AI pilots cannot scale because
40:18
I mean, the technology, it's nothing wrong, nothing
40:21
inherently wrong with the technology, but it's
40:23
just in its maturity phase. It's going through
40:25
that hype cycle right now where where it will
40:28
eventually get mature and it will eventually
40:30
have more use cases. So go for those well -established
40:34
use cases like process mining to understand your
40:37
business processes. And secondly, I mean, I always
40:41
say that if you are in 2026 and you are still
40:44
using Excel or you're still doing manual three
40:50
-way matches. You are you are not in 2026. You
40:53
should be using air and you should be using AI
40:56
for your account receivable account payable They
40:59
are much more there are a lot of established
41:01
solutions out there well tested in the market
41:04
you can you can actually get a lot of uh sort
41:09
of endorsements from from other other industry
41:12
partners as well to understand that okay what
41:14
how did it work or how it did not work so there
41:16
are a lot of feedback available for those uh
41:19
for those tools as well so you can actually actually
41:21
find a good mature solution because these problems
41:24
have been quite they're common across across
41:27
all the verticals mostly across every organization
41:30
has a problem Every organization had bank recon
41:32
problem. So there are quite a good mature solutions
41:35
available out there, which you can then try on
41:37
and use. So go with those in the beginning and
41:40
then more on the on the planning side, go for
41:44
demand forecasting. Again, quite a mature predictive
41:48
analytics capability is out there. People are
41:51
now actually being able to predict their revenues.
41:55
And I bought on something as well a couple of
41:57
a couple of months ago. revenue analytics or
42:01
demand forecasting is another very well established
42:05
and a good use case for AI to for finance to
42:10
start their journey in the beginning. And of
42:12
course, it depends on again, as I said. a caveat
42:15
that it depends on what's your current business
42:17
need. It might not be demand forecasting, but
42:20
having said that, thinking in those directions,
42:23
the principle remains the same that go with the
42:25
traditional AI and think about those traditionalistic,
42:29
more deterministic artificial intelligence. Okay,
42:34
one use case to coming to your second question
42:36
around that use case which didn't work Well,
42:39
I think is is more around and I gave this example
42:42
quite often in my workshops as well is around
42:44
the cash flow forecasting Okay, somehow couple
42:47
of years ago. We were sold when you know before
42:50
generating AI and all that high we were sold
42:52
this this concept that AI is pretty when it takes
42:58
is going to just Use your cat you just forecast
43:01
your cash flow and all your training needs right
43:03
away and you know, you're you're just have your
43:05
cash flow for That I believe is is a bit of and
43:09
I have seen that not working because That is
43:12
bit of an over simplification of the problem
43:15
at hand cash flow at the end of the day is not
43:18
just a number It's a function of your receivables
43:21
payables your funding your financing activities
43:24
operating activities so on so forth It is much
43:27
more complicated than just a simple number. So
43:30
if we are just using historical cash flows to
43:32
predict the future outflow, we are just using
43:35
a statistical model. We are not really using
43:37
predictive analytics in its true sense. What
43:40
we need to do in order for us to make the right
43:44
treasury decisions or cash flow decisions, we
43:46
need to, in fact, be predicting our accounts
43:50
receivable correctly. We must be predicting our
43:53
accounts payable correctly, other cash inflows
43:57
and outflows correctly in order for us to then
44:00
eventually get that cash flow forecasting right.
44:03
I see a lot of times, in fact, in one particular
44:05
case is see, I saw that happen where the organization
44:09
just used a simple historical cash flows to predict
44:12
the future cash flow, which is again, nothing
44:15
better than you know, what I would do as a last
44:19
last 12 months moving average and then just say
44:22
that, you know, this is what it is. There is
44:24
no underlying business driver that you're going
44:27
to be using to predict that cash flow. So these
44:32
are some of the things we need to really think
44:34
about in a little bit more deeper way, unpack
44:37
what that is all about. And again, it comes down
44:39
to that building, again, that digital mindset,
44:42
the ability to critically think and evaluate,
44:44
ask the right questions, and understand what
44:46
is it that we need to really solve and then go
44:50
from there. Wow, thank you so much for sharing.
44:54
And I think sometimes we oversimplify things
44:56
because it's sales versus the actual substance.
45:02
Because I was in the midst of implementing a
45:05
new system and they were like, oh, with this
45:08
new system, it has AI will allow you to do X,
45:10
Y, Z. Yeah. And, and, and the, the, the epic
45:15
one is that where people just look at, just get
45:18
that chat GPT enterprise version or, or copilot
45:21
enterprise version. And they expect that to somehow
45:25
transform their data, somehow transform their
45:28
operations. Yeah. That's not what chat GPT is
45:31
meant to do. It's a very different, it's just
45:33
a sort of a chat pod, which can work in a very
45:37
different way. For your data to work, you need
45:40
to have something else. You need to build data
45:42
pipelines. You need to build data architect,
45:45
right? Data architectures, right? Mindset to
45:47
be able to use that. So again, oversimplifying
45:50
this or seeing the generative AI or agentic AI
45:55
as a magic, as something like a silver bullet,
45:59
that's pretty common. You are absolutely right.
46:01
Yeah. So I'm now curious, how would you finish
46:03
this sentence? The finance teams that will thrive
46:07
with AI are the ones that... Okay, that's a good
46:11
one. The finance teams, do you mind saying that
46:17
again? The finance teams that will thrive. So
46:20
how would you finish this sentence? The finance
46:22
teams that will thrive with AI are the ones that...
46:26
Are the ones that will... re -imagine the fundamental
46:30
of business operating model. Again, as I always
46:35
say, now to unpack this a little bit, we need
46:38
to, that's what even my book re -imagine finance
46:40
is all about, right? We need to re -imagine what
46:43
AI, how AI can help us transform our business
46:47
operations. How do we fundamentally redesign
46:49
our operating models? Because the current operating
46:51
model is not working, is not going to work in
46:54
the new digital or DNA or the... with AI not
47:01
just helping us do work, actually doing the work.
47:05
So you need a redesigned or reimagined business
47:07
operating model. That was a good one. Thank you.
47:13
Thank you so much for sharing. I have a last
47:15
question coming up on time here. So away from
47:18
your work in AI, I'm always curious to hear what's
47:20
your favorite thing to do outside of work? Oh,
47:23
I love I love music. You might just notice that
47:26
there is this. So I like Pakistani classical
47:29
music. So I do a bit of that helps me with refreshing
47:34
my thoughts. And it's quite refreshing, actually.
47:37
Then I'm also I also do a bit of oil portraits.
47:43
I'm a bit of I have this thing around around
47:46
art. So I do portraits as well. And I'm a bit
47:51
of a history buff. In fact, that's why a lot
47:54
of references in my book as well are based on
47:59
human history. So I love that as a subject to
48:04
understand how as humans we evolved, how where
48:07
we are today, how did we get here? That's something
48:14
which fascinates me a lot. And I think a lot
48:16
of that has implication towards where we are
48:19
going. and AI in the picture, how that is going
48:22
to look like. So yeah, this is what I like doing
48:26
when I'm not talking about AI. That is so good.
48:30
And I love history as well, because it reminds
48:33
you that you think you're different, but you're
48:36
really not. You're really not. Yeah, we are overestimating
48:40
our intelligence. See, I mean, we have already
48:46
built things which can actually totally destroy
48:51
humanity, right? So I don't think so. We are
48:54
very smart. We are probably the only, as Yuval
48:59
Noah Harari says, one of my favorite historians,
49:01
he says that, he is my favorite historian, that
49:04
we are the only species who has the ability to
49:08
destroy ourselves fully. But it's so true. Well,
49:16
thank you. Thank you so much, Derek, for being
49:19
on the show. I really, really enjoyed our conversation.
49:23
Same here. Thank you so much, Wasiyah, for having
49:25
me on the show, and I loved our conversation
49:27
as well. Thank you. Thank you. Thank you for
49:30
tuning in to another episode of the Diary of
49:32
a CFO podcast. If you found the conversation
49:35
helpful, please don't forget to leave us a review
49:37
on Apple, Spotify, YouTube, wherever you find
49:40
it, because it really helps getting these kinds
49:42
of insights in front of other finance leaders.
49:45
If you want to go deeper, don't forget to visit
49:47
thediaryofacfo .com for additional resources
49:50
that you can use in your career today.